Saturday, December 10th @ Room 133 + 134
 8:20-8:30 Welcome
Invited Talk 1: The Data-Fusion Problem: Causal Inference and Reinforcement Learning
Elias Bareinboim, Purdue University
Invited Talk 2: A Contextual Research Program
John Langford, Microsoft Research [Slides]
 10:00-11:00 Poster Session 1*, with Coffee Break at 10:30.
 11:00-11:30  Contributed Talk 1: Optimal and Adaptive Off-policy Evaluation in Contextual Bandits 
Yu-Xiang Wang (CMU), Alekh Agarwal (Microsoft) and Miroslav Dudik(Microsoft) [Slides]
 11:30-12:00 Contributed Talk 2:  Joint Causal Inference on Observational and Experimental Datasets
Sara Magliacane (University of Amsterdam), Tom Claassen (Radboud University Nijmegen) and Joris Mooij (University of Amsterdam) [Slides]
 12:00-13:45 Lunch break
 13:45-14:30 Invited Talk 3: Extracting Templates from Media Event Sequences
Marko Grobelnik, Jožef Stefan Institute
 14:30-15:00 Contributed Talk 3: Using Causal Inference to Estimate What-if Outcomes for Targeting Treatments
Qing Liu, Katharine Henry, Yanbo Xu and Suchi Saria (John Hopkins University)
 15:00-16:00 Poster Session 2*, with Coffee Break at 15:00.
 16:00-16:30 Contributed Talk 4: Long-Term Causal Effects in Policy Analysis
Panagiotis (Panos) Toulis (University of Chicago) and David C. Parkes (Harvard University) [Slides]
 16:30-17:15 Invited Talk 4: Causal Inference for Recommendation Systems
David Blei, Columbia University [Slides]
 17:15-18:00 Panel Discussion and Closing

* Each poster will be presented in both sessions.



Invited Talks: Abstracts

Elias Bareinboim, Purdue University: The Data-Fusion Problem: Causal Inference and Reinforcement Learning


John Langford, Microsoft Research: A Contextual Research Program

A theory of contextual interventions has developed and matured to the point where contextual bandits can be routinely deployed to solve appropriate problems.  A more general theory of contextual interventions in complex settings appears desirable and is under development leading to developments in two new areas:

  1. Sequential decision making around deviations from existing solutions  
  2. Global exploration strategies for arbitrary contexts.

Marko Grobelnik, Jožef Stefan Institute: Extracting Templates from Media Event Sequences

In this preliminary research we'll present early results on extracting repeatable probabilistic templates  from global media-event sequences. Such patterns could hint on some weak forms of causality in the global social dynamics. As a basis, we are using the evolving graph of interlinked events generated by the "Event Registry" system (, where each event is represented as an object composed from three main components: social, topical and temporal. In the analysis we will show early results on the structure of the problem and the spectrum of extracted templates from simple to hard ones.

David Blei, Columbia University: Causal Inference for Recommendation Systems

We develop a causal inference approach to recommender systems.  Observational recommendation data contains two sources of information: which items each user decided to look at and which of those items each user liked.  We assume these two types of information come from different models---the exposure data comes from a model by which users discover items to consider; the click data comes from a model by which users decide which items they like.  Traditionally, recommender systems use the click data alone (or ratings data) to infer the user preferences.  But this inference is biased by the exposure data, i.e., that users do not consider each item independently at random.  We use causal inference to correct for this bias. On real-world data, we demonstrate that causal inference for recommender systems leads to improved generalization to new data.

(Joint work with Dawen Liang and Laurent Charlin)